Content
# PaperMCP — All-in-One Academic Paper MCP Server
A unified [Model Context Protocol](https://modelcontextprotocol.io/) server that combines **6 academic paper sources** into a single server, giving AI assistants comprehensive access to research papers.
## 📚 Supported Sources
| Source | Description | Tools |
|--------|-------------|-------|
| **ArXiv** | Search, download & read arXiv papers | 4 |
| **HuggingFace Daily Papers** | Trending ML papers from HuggingFace | 1 |
| **Google Scholar** | Search papers & get author info | 3 |
| **OpenReview** | Conference papers (ICLR, NeurIPS, ICML) | 4 |
| **DBLP** | CS bibliography, BibTeX, fuzzy search | 5 |
| **PapersWithCode** | Papers + code repos + benchmarks | 11 |
| **Unified** | Cross-source search, download, read, list | 4 |
| **Total** | | **32 tools** |
## 🚀 Quick Start
### Install
```bash
# From source
cd PaperMCP
pip install -e .
# Or with uv
uv pip install -e .
```
### Run
```bash
paper-mcp
# or
python -m paper_mcp
```
### MCP Client Configuration
**Claude Desktop / Cursor / Claude Code:**
```json
{
"mcpServers": {
"paper-mcp": {
"command": "paper-mcp",
"env": {
"PAPER_MCP_STORAGE": "~/.paper-mcp/papers"
}
}
}
}
```
> **Have an OpenReview account?** Add your credentials to unlock OpenReview tools:
> ```json
> "env": {
> "PAPER_MCP_STORAGE": "~/.paper-mcp/papers",
> "OPENREVIEW_USERNAME": "your_email@example.com",
> "OPENREVIEW_PASSWORD": "your_password"
> }
> ```
**VS Code (`.vscode/mcp.json` in your workspace):**
```json
{
"servers": {
"paper-mcp": {
"type": "stdio",
"command": "paper-mcp",
"env": {
"PAPER_MCP_STORAGE": "~/.paper-mcp/papers"
}
}
}
}
```
**Claude Code (quick):**
```bash
claude mcp add paper-mcp -- paper-mcp
```
## 🛠️ All Tools
### 🔗 Unified Tools (cross-source)
These 4 high-level tools fan out to **all sources in parallel** and merge results:
| Tool | Description |
|------|-------------|
| `search` | Search across all/selected sources at once. Filter with `sources="arxiv,dblp,scholar"` |
| `download` | Download a paper by ID or URL. Auto-detects source (ArXiv, OpenReview, any PDF URL) |
| `read` | Read a paper's full text. Auto-detects source from ID/URL format |
| `list_papers` | List locally downloaded papers + today's HuggingFace trending papers |
**`search` examples:**
```
search("transformer architecture") # all sources
search("attention mechanism", sources="arxiv,dblp") # specific sources
search("diffusion models", sources="scholar,pwc", max_results=10)
```
**`download` / `read` auto-detection:**
```
download("2401.12345") # ArXiv ID
download("https://arxiv.org/abs/2401.12345") # ArXiv URL
download("https://openreview.net/pdf?id=abc123") # OpenReview URL
read("2401.12345") # reads downloaded ArXiv paper
read("https://example.com/paper.pdf") # reads any PDF URL
```
---
### ArXiv
| Tool | Description |
|------|-------------|
| `arxiv_search` | Search papers with date ranges and category filters |
| `arxiv_download` | Download paper PDF by arXiv ID |
| `arxiv_list` | List all locally downloaded papers |
| `arxiv_read` | Read full text of a downloaded paper |
### HuggingFace Daily Papers
| Tool | Description |
|------|-------------|
| `hf_daily_papers` | Get trending papers for a date (defaults to today) |
### Google Scholar
| Tool | Description |
|------|-------------|
| `scholar_search` | Search papers by keywords |
| `scholar_advanced_search` | Search with author & year filters |
| `scholar_author_info` | Get author profile (affiliation, citations, top pubs) |
### OpenReview
| Tool | Description |
|------|-------------|
| `openreview_search` | Search papers across conferences by keywords |
| `openreview_conference_papers` | Get papers from a specific venue/year |
| `openreview_user_papers` | Get papers by user email |
| `openreview_search_user` | Find user profile by email |
### DBLP
| Tool | Description |
|------|-------------|
| `dblp_search` | Search with boolean queries (AND/OR), year & venue filters |
| `dblp_fuzzy_title_search` | Fuzzy title matching with similarity threshold |
| `dblp_author_publications` | Get author's publications with fuzzy name matching |
| `dblp_venue_info` | Get venue information (journal/conference) |
| `dblp_bibtex` | Fetch BibTeX entry by DBLP key |
### PapersWithCode
| Tool | Description |
|------|-------------|
| `pwc_search_papers` | Search by title, abstract, or ArXiv ID |
| `pwc_get_paper` | Get paper metadata by ID |
| `pwc_paper_repositories` | List code repos for a paper |
| `pwc_paper_datasets` | List datasets for a paper |
| `pwc_paper_methods` | List methods in a paper |
| `pwc_paper_results` | List benchmark results |
| `pwc_search_authors` | Search authors by name |
| `pwc_author_papers` | List papers by author |
| `pwc_conferences` | List/search conferences |
| `pwc_conference_papers` | Papers for a conference proceeding |
| `pwc_read_paper` | Read paper from PDF/HTML URL |
## ⚙️ Environment Variables
| Variable | Description | Required |
|----------|-------------|----------|
| `PAPER_MCP_STORAGE` | Local path for downloaded papers | No (default: `~/.paper-mcp/papers`) |
| `OPENREVIEW_USERNAME` | OpenReview email for API access | No (enables OpenReview tools) |
| `OPENREVIEW_PASSWORD` | OpenReview password | No (enables OpenReview tools) |
| `OPENREVIEW_BASE_URL` | OpenReview API URL | No (default: `https://api2.openreview.net`) |
## 📦 Dependencies
- Python 3.11+
- `mcp` — Model Context Protocol SDK
- `arxiv` — ArXiv API client
- `scholarly` — Google Scholar
- `openreview-py` — OpenReview API
- `httpx` / `requests` — HTTP clients
- `beautifulsoup4` — Web scraping
- `PyPDF2` / `pymupdf4llm` — PDF extraction
## 📝 Example Usage
```
> Search arXiv for recent transformer papers in AI
> Get today's trending papers from HuggingFace
> Find papers by "Yann LeCun" on Google Scholar
> Search ICLR 2024 papers about "token merging" on OpenReview
> Get BibTeX for "conf/nips/VaswaniSPUJGKP17" from DBLP
> Find code repositories for attention papers on PapersWithCode
```
## Credits
Inspired by and learned from these excellent MCP servers:
- [blazickjp/arxiv-mcp-server](https://github.com/blazickjp/arxiv-mcp-server)
- [huangxinping/huggingface-daily-paper-mcp](https://github.com/huangxinping/huggingface-daily-paper-mcp)
- [JackKuo666/Google-Scholar-MCP-Server](https://github.com/JackKuo666/Google-Scholar-MCP-Server)
- [anyakors/openreview-mcp-server](https://github.com/anyakors/openreview-mcp-server)
- [szeider/mcp-dblp](https://github.com/szeider/mcp-dblp)
- [hbg/mcp-paperswithcode](https://github.com/hbg/mcp-paperswithcode)
## License
MIT
